Problem
Equinix's Enterprise Architecture team was building six GenAI pilots before I joined. Each was owned by a different business unit and built without design involvement. Each one worked. But employees had to know which pilot did what, open the right one, and adjust to its interface.
Takeaway
Six pilots that worked in isolation. The system between them didn't.
Reframe
I was asked to bring design consistency across the six pilots and, where possible, bridge them. After the first round of stakeholder conversations, the problem underlying the brief became clear: bridging six internally built pre-production tools would preserve the fragmentation the pilots had created.
I took the reframing to Srivats Ravichandran. We aligned on unifying the pilots into a single company-wide tool with domain routing underneath, then pitched it together to stakeholders across Sales, Human Resources (HR), Operations (Ops), Information Technology (IT), Legal, and Enterprise Architecture (EA).
Takeaway
The pitch landed. Budget and ownership didn't. A decision on an all-employee tool couldn't come from any one of the six business units. It had to come from executive decision-makers leading Equinix's GenAI initiatives, and nothing on the table was moving them.
Challenge
Each of the six business units had reasons to protect its own pilot. Competing priorities, fragmented decision-making, and the lack of centralized direction for an all-employee tool meant that a unified product couldn't be negotiated through meetings alone.
Stakeholders needed evidence, not a vision deck. I built the unified experience as a working prototype, put it in front of real users across regions, and let the research data do the alignment work the pitch couldn't.
Takeaway
The validation wasn't a post-launch check. It was how alignment and funding were earned.
Design
Intervention 1
Ran stakeholder design workshops to establish shared vocabulary.
Jung-Soo Choi, Sal Sidani, and I co-facilitated three workshops across Sales, Operations, and IT. We used Jobs-to-be-Done to align on what employees needed from a GenAI tool, and Killer Questions to surface assumptions and risks across each business unit. The workshops produced the first shared problem statement across the teams.
Intervention 2
Designed domain routing as invisible infrastructure.
I designed the user-facing interaction model. Srivats engineered the routing underneath. Sales queries went to the sales pilot. HR queries to HR. Multi-domain questions pulled from multiple APIs, with information consolidated into a single response. The six pilots stayed alive underneath as domain APIs.
Intervention 3
Built a custom design system for AI-native interaction patterns.
Sal Sidani and I built a design system for conversational AI. I created AI-native components that the existing library didn't cover. Sal reviewed for consistency against the broader Equinix visual language. These patterns needed their own foundation, built once.
Intervention 4
Added guided prompts to solve the blank-page problem.
Early usability sessions showed that employees who hadn't used GenAI tools before didn't know how to start. A blinking cursor produced hesitation. I designed guided prompts: role-based, activity-based, and generic starters. Users could edit, ignore, or send them.
Intervention 5
Ran sentiment analysis as validation and alignment.
I designed a mixed-methods study, conducted with six active beta users of E-GPT across EMEA, APAC, and AMER. Jung-Soo and I co-ran interviews, combining System Usability Scale (SUS) scoring before and after the redesign with open-ended questions on satisfaction, value, and usage. I translated findings into design revisions.
